Speech compression is used to compress the data of digital audio signals possessing speech. Speech compression utilizes evaluation of speech-specific limitations, using processing techniques for audio signals to replicate the speech signal, united with the general data compression algorithms to symbolize the resultant modeled restrictions in a solid bitstream.
The 2 most significant uses of speech compression are Voice over IP and mobile telephony.
Speech compression varies from all other kinds of audio compression because in speech compression speech comes as an uncomplicated signal than most of the other audio signals, and a lot of statistical information is presented about the characteristics of speech. Because of this, some aural info which is applicable in audio compression becomes pointless in speech compression. In speech compression, the most significant principle is conservation of transparency and "loveliness" of speech, with a controlled quantity of transmitted data.
Techniques of speech compression:
Most of the efforts in the field of speech compression were aggravated by military research of digital communications for safe military radios, in which low data rates were needed to permit efficient action in an aggressive radio atmosphere. Lot more power for processing was made available in the shape of VLSI integrated circuits. Because of this, contemporary speech compression algorithms can use more composite techniques.
The most commonly used speech compression system is CELP (Code Excited Linear Prediction) compression, which is utilized in the GSM standard. In CELP, the process of modeling is split into 2 stages. The first one is a linear analytical stage that molds the ethereal envelope and code-book based form of the residue of the linear analytical model.